CMPSTheory & User Reference Manual
Model Coupling and Recommended Use
LIKUA HomeManual Home

Model Coupling and Compatibility

This page summarizes how the user-selectable CMPS model families add equations and which combinations are available together. Model availability is combination-dependent: two individually available controls are not necessarily supported as a combined formulation.

Coupled equation ordering

The complete primitive block is ordered by active model family:

\[\underbrace{\mathbf V_f}_{\text{carrier flow}},\quad \underbrace{\mathbf V_t}_{\text{turbulence}},\quad \underbrace{\mathbf V_d}_{\text{dilute dispersed phase}},\quad \underbrace{\mathbf Y}_{\text{species}},\quad \underbrace{\boldsymbol\alpha}_{\text{VOF}}.\]

Inactive model blocks are omitted. VOF, when enabled, occupies the final block. Solid cells use only the temperature equation even though the distributed matrix retains the common block layout.

Automatic compressible/incompressible selection

The active material EOS set is authoritative. Without VOF, the carrier set must be uniformly constant-density or uniformly pressure-dependent. With VOF, mixed EOS constituents are allowed: all constant-density constituents select artificial compressibility; any pressure-dependent constituent selects the compressible formulation. In the GUI this regime is reported as Automatic by default rather than exposed as an independent physics switch; formulation-specific controls follow the resulting regime.

VOF compatibility

CapabilityCurrent VOF status
Steady UpwindSupported
Steady HRICSupported; no pseudo-time Courant substitution
Transient Upwind / HRICSupported
Transient CICSAMSupported
Compressible mixed EOSSupported with AUSM carrier flux
All-constant-density mixtureSupported with artificial compressibility
MPI domain decompositionSupported with VOF partition consistency checks
Agglomeration multigridCurrently locked for VOF
Split volumetric reaction integrationCurrently locked for VOF
Non-adiabatic VOF walls / VOF CHTCurrently locked

VOF versus dilute dispersed phase

These are separate multiphase formulations. VOF represents resolved or macroscopically separated constituents sharing a common velocity, pressure and temperature. The dilute dispersed-phase model transports a separate particle/droplet velocity, dispersed density, temperature and optional interfacial area with explicit momentum/heat exchange to the carrier.

Species and reactions

Species transport is part of the carrier coupled block and may be combined with laminar or turbulent finite-rate reaction models in the non-VOF path. The current CMPS formulation explicitly rejects the split volumetric reaction operator when VOF is active.

Solid and charring coupling

Single-carrier flow can couple to solid conduction through fluid–solid walls, including non-adiabatic and charring variants. Homogeneous VOF currently uses adiabatic wall compatibility only.

Rotating-reference-frame compatibility

Rotating zones use grid-relative mass transport with absolute velocity/energy. Compressible rotating carrier flow currently requires AUSM. The dilute dispersed phase is rejected with rotating zones. The VOF path uses the homogeneous carrier transport and its dedicated boundary fluxes; it remains subject to the general compressible-AUSM requirement and the VOF-specific locks listed above.

Selectable versus non-selectable models

The current selectable carrier-flow model set is inviscid, laminar, and GE k–omega. SAS is an optional extension of the GE turbulence path. The GUI presents the turbulence models that are currently selectable for production cases; options not shown in the Physics panel are outside the user-selectable model set described here. Likewise, an exposed shock-sensor option is not promoted here as an active flux feature where no numerical consumption of that selector is found in the current solve path.

NeuralFlow learned-solver compatibility

The current learned state is exactly \([u_1,\ldots,u_D,p,T]\). Training is serial-only. The learned boundary schema explicitly supports far field, pressure outlet, ordinary wall and symmetry faces; one-dimensional charring and one-sided interior/FSI-wall graph edges are rejected. VOF and any equation layout that enlarges the primitive block beyond \(D+2\) run through the conventional CFD path. See NeuralFlow: Solver-Trained Recurrent GNO.